Home security intelligent monitoring or turning point

Home security is inseparable from the video surveillance system. The video surveillance system has also been widely used in community security monitoring, fire monitoring, traffic violations, flow control, military, and security in public places such as banks, shopping malls, subways, and airports. There will be wider application prospects.

However, existing video surveillance systems can only rely on humans to complete a series of analysis and processing such as monitoring, alarming, and recording. They do not take full advantage of real-time active monitoring. More product requirements and broader market applications require intelligent analysis and processing capabilities in future security products. From the current level of technology, digital image processing technology has been increasingly mature, artificial intelligence has also been making rapid progress. How to integrate the functions of image intelligent analysis into smart products has become the focus of various companies and manufacturers. An intelligent video surveillance system with advanced intelligent analysis not only greatly enhances the ability to monitor and reduce hidden dangers, but also saves manpower and material resources and reduces costs.

The development of intelligence in the monitoring field Before the early 1990s, closed-circuit television monitoring systems based on analog devices were the earliest first-generation monitoring systems. The technology was mature and stable, and it was widely used in many practical projects. For some applications, this technology is still in the eyes of people. The true intelligent monitoring comes along with the development of the computer and the innovation of the network. In the mid-1990s, the improvement of computer processing capabilities and the development of video technology enabled people to start using the computer's high-speed data processing capabilities to capture and process video. The resulting digital monitoring system was called the second generation of video surveillance systems. . Since the late 1990s, with the increase of network bandwidth, computer processing capacity and storage capacity, and the emergence of various video processing technologies, video surveillance has entered a new era of digitalization. The application of intelligence in security is increasingly important. Therefore, the third generation monitoring system is based on the network, with digital video compression, transmission, storage and playback as the core, and a technological revolution featuring intelligent image analysis. Into the police, illegal stay, items left behind, moving target tracking and so on. These advanced intelligent analysis functions have become a major feature of modern security monitoring products.

The study of related intelligent algorithms mentioned that security cannot fail to mention intelligence, mentioning that intelligence has to mention algorithms. The algorithm, as the soul of behavior analysis and image processing, plays a guiding role in future product development and program implementation. The continuous expansion of the monitoring field and the continuous development of the market have led to huge video information analysis and processing. For intelligent video surveillance systems, facing such a large amount of information, good technologies and structures are the key issues. In the process of remote monitoring, in order to reduce the pressure of network transmission, some intelligent analysis modules are embedded in the front end of the device into the encoding module of the product, and only the useful video information is extracted through the previous intelligent analysis and the compression and transmission thereof is performed. , greatly reducing the amount of data transmitted, and the image analysis part is processed by the software part of the back end after decoding. This division of labor is much better than simple centralized software or hardware processing.

Compared with H264 and other codec algorithms, the reason why video intelligence algorithm does not have a unified standard, without ASIC, is because of its complexity and diversity of target behavior, it is difficult to develop a set of standard rules to apply to each industry sector. Intelligent algorithm belongs to the emerging comprehensive science technology. Many effective intelligent analysis algorithms are first developed from abroad, and there are still a considerable part of the algorithm is either too high complexity, not suitable for real-time requirements; or poor adaptation The dependence on the scene is too great; either it still stays at the stage of theoretical research and is still far from the product. There are few intelligent surveillance products that have truly independent intellectual property rights in China, and the vast majority of products are derived from developed countries such as the United States, Europe, and Israel. Most of the domestic production and installation of video surveillance systems still remain in non-intelligent areas. The "smart video surveillance" mentioned in the surveillance system actually stays in the concept of ordinary network video surveillance.

Most of them are IP surveillance and digital surveillance. They do not have the content of advanced intelligent analysis. For this reason, many companies in China have also invested a large amount of human and material resources in the area of ​​intelligence, and have made a lot of improvements in the traditional detection of moving targets and achieved breakthrough results. The intelligent security products with autonomous knowledge products play a very important role in improving the overall intelligent monitoring level in China and improving the status of domestic products in international competition.

From an algorithmic point of view, there are many researches to be done on the technical aspects of monitoring products. However, this also gives us an opportunity. Whoever masters the core technologies in the future development will be able to dig into the A barrel of gold, occupying the market in related fields. In intelligent image analysis, the detection and tracking of moving objects should be the most basic function in advanced intelligent analysis. It is the basis for the analysis of behaviors such as intrusion alarms, illegal stays, and lost relics. The following is a brief analysis of the process of object recognition tracking.

Specific application examples The implementation of the algorithm can be seen as a modeling process, but also as an applied mathematics project. Detection and recognition of moving objects means that separating the changing foreground from the background image in the sequence image is a major part of digital image processing. Changes in weather and lighting, interference from other objects in complex backgrounds, motion shadows, camera motion, and jitter all make it difficult to extract the target. Therefore, accurate detection and accurate tracking of moving targets have become an important technical point in video surveillance systems. The whole process can be simply divided into the following steps: video preprocessing, target detection and recognition, target classification, target tracking, behavior analysis and rule creation.

The video preprocessing video preprocessing process actually reduces noise and highlights the tracking target. There are two types of image noise, one is source noise, and the other is observation noise. Changes in ambient light, background leaves, rain and snow are all sources of noise; observation noise mainly refers to the camera jitter, transmission line interference and other effects. These noises can cause troubles in tracking the target extraction process. The commonly used methods for eliminating noise are digital image stabilization, background adaptive learning, threshold segmentation, morphological filtering, and binarization. These processes can eliminate most of the noise in the image and reduce its interference with the target detection.

Target Detection and Recognition After using the moving target extraction process to obtain foreground moving objects, it is necessary to have an identification process for the target. This recognition is divided into single-target manual recognition and multi-target recognition. For individual object tracking, the object of interest can be artificially selected from the video and tracked. For areas of multiple objects, an area can be divided to identify only the objects entering the area. track.

Target classification does not require classification for individual object tracking within the surveillance area. Even if the object is obstructed during the movement, when the object satisfies the linear motion and the background changes relatively smoothly, the algorithm can still accurately track the object. For non-linearly moving objects in complex environments, we add some target feature information to help identify the tracking based on clinic-to-single-target tracking. Target classification is generally applied to tracking single goals in multi-target situations, or multi-target tracking multi-target environments. When there are multiple moving targets in the environment, we can extract the targets of each motion and classify them according to the characteristics such as position, velocity, shape, texture, and color. First, save the feature information of the target of interest and then track it. When the tracked object disappears after the video disappears, we use the previously saved feature information to match all the moving objects that appear in the current frame. By comparing the results, it can be determined which one of the objects is the one just tracked.

Behavior analysis and rule creation If the tracking process of an object is a purely technical process, behavior analysis should belong to the category of advanced intelligence. Simply speaking, behavior analysis is based on the creation of artificial rules, the process of automatically analyzing the images in the video and extracting the key information in the video source. Because the creation of rules is related to the environment, such as the situation in the scene, setting the conditions and areas for triggering the alarm is also different. Therefore, in the development of intelligent security, the focus of software-side work is to make intelligent analysis as far as possible in the direction of humanization. Under simple settings, intelligent analysis can allow the entire image to have a process of self-regulated learning and further reduce the number of people. The influence of the factors reduces the workload of people.

Conclusions and Outlook With the rapid development of China's security industry and the continuous expansion of the market size, more and more companies are engaged in the security industry. For companies that promote smart security products, only by grasping the pulse of the security market and mastering key technologies can they correctly respond to the needs of the future market and continuously introduce new ones, constantly discovering opportunities in the market competition.

Compared with foreign countries, China’s smart security technology still has a long way to go, but it can be said with certainty that these gaps will increase with the frequent increase in the number of domestic and international technological exchanges, the doubling of the number of domestic talents, and the improvement of the R&D environment. The smaller it comes. It is hoped that in the future, various companies engaged in the security industry in the country will also be able to communicate and exchange information on the premise of protecting the company’s independent intellectual property rights, and will continue to improve the product specifications and market specifications of the security industry. Under the premise of establishing a win-win market, continuous technological innovation has spread intelligent video surveillance products with intelligent image processing technology as the core in various industries and fields, creating a peaceful and safe environment for people's lives.